Repository navigation
Publish the embedded SDK boundary for ReviewBot - #115
Conversation
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 0e9ea957e1
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| page_text = self._document_path(page).read_text( | ||
| encoding="utf-8" | ||
| ) | ||
| if self._rule_exists(page_text, rule_id): | ||
| reason = "rule_id already exists in the target page" |
There was a problem hiding this comment.
When model output restates an existing rule under a new valid rule_id, this check accepts it because _rule_exists compares only headings; apply() then appends the synonymous section. That directly bypasses the repository's union-first requirement and allows repeated curation runs to accumulate duplicate conclusions instead of merging them.
AGENTS.md reference: AGENTS.md:L7-L11
Useful? React with 👍 / 👎.
| _VALIDATOR_IDS: Final = ( | ||
| "knowledge/tools/check_knowledge_tree.py", | ||
| "knowledge/tools/check_wiki_lint.py", | ||
| ) |
There was a problem hiding this comment.
For any accepted proposal targeting knowledge/repos/vllm-omni/, apply() runs only these two validators and can return success=True without the required tools/audit_vllm_omni_release.py enforce check. The two validators do not cover owner-document, SHA-pin, or upstream-source reconciliation, so the curator can approve a batch that later fails the mandatory PR audit.
AGENTS.md reference: AGENTS.md:L25-L29
Useful? React with 👍 / 👎.
| @lru_cache(maxsize=8) | ||
| def _resource_revision(knowledge_dir: str, adapters_dir: str) -> str: |
There was a problem hiding this comment.
In an editable/source deployment, if knowledge or adapter files change after plan() issues a context, this path-keyed cache keeps returning the original digest indefinitely. Consequently the resource-drift check in DirectClient.validate() can still mark the review complete even though its referenced provider resources changed; the revision must be refreshed or the cache invalidated before that comparison.
Useful? React with 👍 / 👎.
Summary
Validation
mcp<2environment: SDK, contract, MCP, and thin-server tests pass